Network Switching (Part VI)

The Rule of the Network Road


Network administrators and designers have traditionally strived to design networks using the 80/20 rule. Using this rule, a network designer would try to design a network in which 80 percent of the traffic stayed on local segments and 20 percent of the traffic went on the network backbone.This was an effective design during the early days of networking, when the majority of LANs were departmental and most traffic was destined for data that resided on the local servers. However, it is not a good design in today’s environment, where the majority of traffic is destined for enterprise servers or the Internet.A switch’s ability to create multiple data paths and provide swift, low−latency connections allows network administrators to permit up to 80 percent of the traffic on the backbone without causing a massive overload of the network. This ability allows for the introduction of many bandwidth−intensive uses, such as network video, video conferencing, and voice communications.Multimedia and video applications can demand as much as 1.5Mbps or more of continuous bandwidth. In a typical environment, users can rarely obtain this bandwidth if they share an average 10Mbps network with dozens of other people. The video will also look jerky if the data rate is not sustained. In order to support this application, a means of providing greater throughput is needed. The ability of switches to provide dedicated bandwidth at wire−speed meets this need.


Switched Ethernet Innovations


Around 1990, many vendors offered popular devices known as intelligent multiport bridges; the first known usage of the term switch was the Etherswitch, which Kalpana brought to the market in 1990. At the time, these devices were used mainly to connect multiple segments—they usually did very little to improve performance other than the inherent benefits bridges provide, such as filtering and broadcast suppression. Kalpana changed that by positioning its devices as performance enhancers. A number of important features made the Kalpana switches popular, such as using multiple transmission paths for network stations and cut−through switching.Cut−through switching reduced the delay problems associated with standard bridges by providing the means to have multiple transmissions paths to network devices. Each device could have its own data path to the switch and did not need to be in a shared environment.Kalpana was able to do this by dedicating one pair of the station wiring to transmitting data and one pair to receiving data. This improvement allowed the Kalpana designers to ignore the constraints of collision detection and carrier sense, because the cables were dedicated to one station. Kalpana continued its history of innovation with the introduction in 1993 of full−duplex Ethernet.

Full−Duplex Ethernet

Prior to the introduction of full−duplex (FDX) Ethernet, Ethernet stations could either transmit or receive data; they could not do both at the same time, because there was no way to ensure a collision−free environment. This was known as half−duplex (HDX) operation.FDX has been a feature of WANs for years, but only the advent of advances in LAN switching technology made it practical to now consider FDX on the LAN. In FDX operation, both the transmission and reception paths can be used simultaneously. Because FDX operation uses a dedicated link, there are no collisions, which greatly simplifies the MAC protocol. Some slight modifications in the way the packet header is formatted enable FDX to maintain compatibility with HDX Ethernet.You don’t need to replace the wiring in a 10BaseT network, because FDX operation runs on the same two−pair wiring used by 10BaseT. It simultaneously uses one pair for transmission and another pair for reception. A switched connection has only two stations: the station itself and the switch port. This setup makes simultaneous transmission possible and has the net effect of doubling a 10Mbps LAN.This last point is an important one. In theory, FDX operation can provide double the bandwidth of HDX operation, giving 10Mbps speeds in each direction. However, achieving this speed would require that the two stations have a constant flow of data and that the applications themselves would benefit from a two−way data flow. FDX links are extremely beneficial in connecting switches to each other. If there were servers on both sides of the link between switches, the traffic between switches would tend to be more symmetrical.



Fast Ethernet


Another early innovation in the switching industry was the development of Fast Ethernet. Ethernet as a technology has been around since the early 1970s, but by the early 1990s its popularity began to wane. Competing technologies such as FDDI running at 100Mbps showed signs of overtaking Ethernet as a de facto standard, especially for high−speed backbones.Grand Junction, a company founded by many of the early Ethernet pioneers, proposed a new Ethernet technology that would run at 10 times the 10Mbps speed of Ethernet. They were joined by most of the top networking companies—with the exception of Hewlett−Packard (HP), which had a competing product. HP’s product, known as 100Mbps VG/AnyLAN, was in most respects far superior to the product proposed by Grand Junction. It had a fatal flaw, though: It was incompatible with existing Ethernet standards and was notbackward compatible to most of the equipment in use at the time. Although the standards bodies debated the merits of each of the camps, the marketplace decided for them. Fast Ethernet is the overwhelming winner, so much so that even HP sells Fast Ethernet on almost all its products.Note In 1995, Cisco purchased both Kalpana and Grand Junction and incorporated their innovations into its hardware. These devices became the Catalyst line of Cisco products.


Gigabit Ethernet


In order to implement Gigabit Ethernet (GE), the CSMA/CD method was changed slightly to maintain a 200−meter collision diameter at gigabit−per−second data rates. This slight modification prevented Ethernet packets from completing transmission before the transmitting station sensed a collision, which would violate the CSMA/CD rule. GE maintains a packet length of 64 bytes, but provides additional modifications to the Ethernet specification.The minimum CSMA/CD carrier time and the Ethernet slot time have been extended from 64 bytes to 512 bytes. Also, packets smaller than 512 bytes have an extra carrier extension added to them. These changes, which can impact the performance of small packets, have been offset by implementing a feature called packet bursting, which allows servers, switches, and other devices to deliver bursts of small packets in order to utilize the available bandwidth.Because it follows the same form, fit, and function as its 10− and 100Mbps predecessors, GE can be integrated seamlessly into existing Ethernet and Fast Ethernet networks using LAN switches or routers to adapt between the different physical line speeds. Because GE is Ethernet, only faster, network managers will find the migration from Fast Ethernet to Gigabit Ethernet to be as smooth as the migration from Ethernet to Fast Ethernet.



Avoiding Fork−Lift Upgrades


Although dedicated switch connections provide the maximum benefits for network users, you don’t want to get stuck with fork−lift upgrades. In a fork−lift upgrade, you pay more to upgrade your computer or networking equipment than it would cost to buy the equipment already installed. The vendor knows that you are not going to buy all new equipment, so the vendor sells you the upgrade at an enormous price. In order to exchange it for the bigger, better, faster equipment It may sometimes be necessary to support legacy equipment.Fortunately for Ethernet switches you can provide connectivity in a number of ways. You can attach shared hubs to any port on the switch in the same manner that you connect end stations. Doing so makes for a larger collision domain, but you avoid paying the high costs of upgrades.Typically, your goal would be to migrate toward single−station segments as bandwidth demands increase. This migration will provide you with the increased bandwidth you need without wholesale replacement of existing equipment or cabling. In this lower cost setup, a backbone switch is created in which each port is attached to the now−larger collision domain or segment. This switch replaces existing connections to routers or bridges and provides communication between each of the shared segments.